Domain-Specific Intermediate Training for Zero-Shot Cross-Lingual Transfer in 1B and 10B Parameter Models on XTREME Tasks

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuning again on the target task---often improves model performance substantially on language understanding tasks in monolingual English settings. We investigate whether English intermediate-task training is still helpful on non-English target tasks. Using nine intermediate language-understanding tasks, we evaluate intermediate-task transfer in a zero-shot cross-lingual setting on the XTREME benchmark. We see large improvements from intermediate training on the BUCC and Tatoeba sentence retrieval tas

Research goal: What is the impact of domain-specific intermediate-task training (e.g., legal vs. biomedical tasks) on zero-shot cross-lingual transfer performance for 1B vs. 10B parameter models, as evaluated by accuracy on XTREME-POS and XTREME-NER tasks?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.3/10.